The bottom line: AI agents require a structured data foundation that precisely maps physical, logical and virtual layers—standard data platforms and CMDBs alone are insufficient.
AI agents can automate operational tasks in IT and network infrastructure, but often fail due to outdated or incomplete operational data. Without precise “ground truth,” they make fast but erroneous decisions.
AI agents promise to significantly increase automation of operational tasks in IT and network infrastructure beyond current systems. They evaluate data, trigger tools, and execute multi-stage workflows with reduced manual effort—from automated capacity planning through dependency tracking to autonomous troubleshooting. An agent could, for example, assess available capacity across multiple sites, propose and execute resource allocations, or trace an application’s dependency chain down to the physical cable segment in seconds.
The central hurdle is what’s known as grounding: the process by which AI algorithms must be linked to precise, current data from the real world. The risk does not arise from fabricated content, but from outdated, incomplete, or unvalidated operational data. A typical scenario: An AI agent plans GPU cluster expansion and identifies free racks. But these were physically reconfigured months ago—without documentation. The agent rapidly initiates erroneous workflows, leading to faulty change tickets, incorrect reservations, or risks to power supply and cooling.
Many organizations rely on central data platforms such as data lakes or CMDBs (Configuration Management Databases). These alone, however, are insufficient. Data lakes are optimized for tabular data but often lack a domain model that maps relationships between ports, cables, circuits, racks, services, and sites. Traditional CMDBs focus on service management and hit their limits at the logical level—the AI agent lacks physical precision, such as which server is connected to which port via which circuit.
Infrastructure data presents special requirements: it is highly relational, changes continuously, and must be regularly validated against physical reality. For AI agents to function reliably, a digital twin model is therefore necessary that correctly maps physical, logical, and virtual layers and serves as a validated data foundation.
Regulatory pressure further tightens requirements: the EU NIS2 Directive mandates operators of critical infrastructure to maintain robust asset and risk management. The Digital Operational Resilience Act (DORA) and the EU AI Act impose additional requirements on transparency and verifiability of AI-based systems.
Source: www.it-daily.net · Published July 8, 2026
Lumi AI News — AI-assisted curation pursuant to Art. 50 EU AI Act. Paraphrase and classification by Lumi News Pipeline v1.7.3.